A Link between Machine Learning and Optimization in Ground-Motion Model Development: Weighted Mixed-Effects Regression with Data-Driven Probabilistic Earthquake Classification

2020 ◽  
Vol 110 (6) ◽  
pp. 2777-2800
Author(s):  
Sebastian von Specht ◽  
Fabrice Cotton
Keyword(s):  
Ground Motion ◽  
Mixed Effects ◽  
Mixed Effects Model ◽  
Data Driven ◽  
Weighted Likelihood ◽  
Motion Data ◽  
Motion Models ◽  
Uncertainty Estimates ◽  
Data Weighting ◽  
Ground Motion Models

ABSTRACT The steady increase of ground-motion data not only allows new possibilities but also comes with new challenges in the development of ground-motion models (GMMs). Data classification techniques (e.g., cluster analysis) do not only produce deterministic classifications but also probabilistic classifications (e.g., probabilities for each datum to belong to a given class or cluster). One challenge is the integration of such continuous classification in regressions for GMM development such as the widely used mixed-effects model. We address this issue by introducing an extension of the mixed-effects model to incorporate data weighting. The parameter estimation of the mixed-effects model, that is, fixed-effects coefficients of the GMMs and the random-effects variances, are based on the weighted likelihood function, which also provides analytic uncertainty estimates. The data weighting permits for earthquake classification beyond the classical, expert-driven, binary classification based, for example, on event depth, distance to trench, style of faulting, and fault dip angle. We apply Angular Classification with Expectation–maximization, an algorithm to identify clusters of nodal planes from focal mechanisms to differentiate between, for example, interface- and intraslab-type events. Classification is continuous, that is, no event belongs completely to one class, which is taken into account in the ground-motion modeling. The theoretical framework described in this article allows for a fully automatic calibration of ground-motion models using large databases with automated classification and processing of earthquake and ground-motion data. As an example, we developed a GMM on the basis of the GMM by Montalva et al. (2017) with data from the strong-motion flat file of Bastías and Montalva (2016) with ∼2400 records from 319 events in the Chilean subduction zone. Our GMM with the data-driven classification is comparable to the expert-classification-based model. Furthermore, the model shows temporal variations of the between-event residuals before and after large earthquakes in the region.

Ground-Motion Prediction Model Based on Neural Networks to Extract Site Properties from Observational Records

2021 ◽  
Author(s):  
Tomohisa Okazaki ◽  
Nobuyuki Morikawa ◽  
Asako Iwaki ◽  
Hiroyuki Fujiwara ◽  
Tomoharu Iwata ◽  
...  
Keyword(s):  
Ground Motion ◽  
Ground Acceleration ◽  
Site Specific ◽  
Motion Data ◽  
Motion Models ◽  
Proposed Model ◽  
Single Station ◽  
Input Variables ◽  
Ground Condition ◽  
Ground Motion Models

ABSTRACT Choosing the method for inputting site conditions is critical in reducing the uncertainty of empirical ground-motion models (GMMs). We apply a neural network (NN) to construct a GMM of peak ground acceleration that extracts site properties from ground-motion data instead of referring to ground condition variables given for each site. A key structure of the model is one-hot representations of the site ID, that is, specifying the collection site of each ground-motion record by preparing input variables corresponding to all observation sites. This representation makes the best use of the flexibility of NN to obtain site-specific properties while avoiding overfitting at sites where a small number of strong motions have been recorded. The proposed model exhibits accurate and robust estimations among several compared models in different aspects, including data-poor sites and strong motions from large earthquakes. This model is expected to derive a single-station sigma that evaluates the residual uncertainty under the specification of estimation sites. The proposed NN structure of one-hot representations would serve as a standard ingredient for constructing site-specific GMMs in general regions.

Conditional Ground-Motion Models for Horizontal Peak Ground Displacement for Active Crustal Regions

2021 ◽  
Author(s):  
Chih-Hsuan Sung ◽  
Norman A. Abrahamson ◽  
Jyun-Yan Huang
Keyword(s):  
Ground Motion ◽  
Earthquake Engineering ◽  
Input Parameter ◽  
Strong Motion ◽  
Tectonic Deformation ◽  
Ground Displacement ◽  
Strong Motion Data ◽  
Motion Data ◽  
Motion Models ◽  
Ground Motion Models

ABSTRACT Ground-motion models (GMMs) are developed for peak ground displacement (PGD) and for bandlimited PGD based on strong-motion data that has been filtered as part of standard processing and the total PGD that includes the tectonic deformation as well as the vibratory ground motion. For the bandlimited PGD, we develop conditional ground-motion models (CGMMs) using subsets of the Pacific Earthquake Engineering Research Center Next Generation Attenuation-West2 Project (NGA-W2) database and the National Center for Research on Earthquake Engineering Taiwan Senior Seismic Hazard Analysis Committee level 3 project database. The CGMM approach includes the observed pseudospectral acceleration (PSA(T)) as an input parameter in addition to magnitude and distance. The period of the PSA(T) is used as an input parameter; it is magnitude dependent and is based on the period for which there is the highest correlation between the ln(PGD) and ln(PSA(T)). Two CGMMs are developed: a global model based on the NGA-W2 data and a region-specific model for Taiwan. The conditional PGD models are combined with traditional GMMs for PSA(T) values to develop GMMs for both the median and standard deviation of PGD without the dependence on PSA. A second set of PGD GMMs are developed to correct for two factors: the effect of the high-pass filtering from standard record processing and the stronger large magnitude (M>6.5) scaling due to tectonic deformation. For magnitudes greater than 7, the PGD values from the total PGD GMMs are 2–5 times larger than the bandlimited PGD values based on the strong-motion data sets, but the increase is at very long periods. The appropriate PGD model to use, bandlimited PGD or total PGD, depends on the period range of interest for the specific engineering application.

Ranking and Selection of Earthquake Ground-Motion Models Using the Stochastic Area Metric

2021 ◽  
Author(s):  
Jaleena Sunny ◽  
Marco De Angelis ◽  
Benjamin Edwards
Keyword(s):  
Ground Motion ◽  
Hazard Analysis ◽  
Cumulative Distribution ◽  
Earthquake Ground Motion ◽  
Motion Data ◽  
Motion Models ◽  
Area Metric ◽  
Few Data ◽  
Ground Motion Models ◽  
Selection Of

Abstract We introduce the cumulative-distribution-based area metric (AM)—also known as stochastic AM—as a scoring metric for earthquake ground-motion models (GMMs). The AM quantitatively informs the user of the degree to which observed or test data fit with a given model, providing a rankable absolute measure of misfit. The AM considers underlying data distributions and model uncertainties without any assumption of form. We apply this metric, along with existing testing methods, to four GMMs in order to test their performance using earthquake ground-motion data from the Preston New Road (United Kingdom) induced seismicity sequences in 2018 and 2019. An advantage of the proposed approach is its applicability to sparse datasets. We, therefore, focus on the ranking of models for discrete ranges of magnitude and distance, some of which have few data points. The variable performance of models in different ranges of the data reveals the importance of considering alternative models. We extend the ranking of GMMs through analysis of intermodel variations of the candidate models over different ranges of magnitude and distance using the AM. We find the intermodel AM can be a useful tool for selection of models for the logic-tree framework in seismic-hazard analysis. Overall, the AM is shown to be efficient and robust in the process of selection and ranking of GMMs for various applications, particularly for sparse and small-sized datasets.

A Suite of Alternative Ground-Motion Models (GMMs) for Israel

2021 ◽  
Author(s):  
Soumya Kanti Maiti ◽  
Gony Yagoda-Biran ◽  
Ronnie Kamai
Keyword(s):  
Seismic Hazard ◽  
Ground Motion ◽  
Empirical Data ◽  
Epistemic Uncertainty ◽  
Ground Motions ◽  
Plate Boundary ◽  
Strong Motion ◽  
Engineering Applications ◽  
Motion Models ◽  
Ground Motion Models

ABSTRACT Models for estimating earthquake ground motions are a key component in seismic hazard analysis. In data-rich regions, these models are mostly empirical, relying on the ever-increasing ground-motion databases. However, in areas in which strong-motion data are scarce, other approaches for ground-motion estimates are sought, including, but not limited to, the use of simulations to replace empirical data. In Israel, despite a clear seismic hazard posed by the active plate boundary on its eastern border, the instrumental record is sparse and poor, leading to the use of global models for hazard estimation in the building code and all other engineering applications. In this study, we develop a suite of alternative ground-motion models for Israel, based on an empirical database from Israel as well as on four data-calibrated synthetic databases. Two host models are used to constrain model behavior, such that the epistemic uncertainty is captured and characterized. Despite the lack of empirical data at large magnitudes and short distances, constraints based on the host models or on the physical grounds provided by simulations ensure these models are appropriate for engineering applications. The models presented herein are cast in terms of the Fourier amplitude spectra, which is a linear, physical representation of ground motions. The models are suitable for shallow crustal earthquakes; they include an estimate of the median and the aleatory variability, and are applicable in the magnitude range of 3–8 and distance range of 1–300 km.

Testing non-linear amplification factors used in ground motion models

2021 ◽  
Author(s):  
Karina Loviknes ◽  
Danijel Schorlemmer ◽  
Fabrice Cotton ◽  
Sreeram Reddy Kotha
Keyword(s):  
Ground Motion ◽  
Site Effects ◽  
Ground Motions ◽  
Site Amplification ◽  
Building Codes ◽  
Linear Amplification ◽  
Motion Models ◽  
Earthquake Predictability ◽  
Non Linear ◽  
Ground Motion Models

<p>Non-linear site effects are mainly expected for strong ground motions and sites with soft soils and more recent ground-motion models (GMM) have started to include such effects. Observations in this range are, however, sparse, and most non-linear site amplification models are therefore partly or fully based on numerical simulations. We develop a framework for testing of non-linear site amplification models using data from the comprehensive Kiban-Kyoshin network in Japan. The test is reproducible, following the vision of the Collaboratory for the Study of Earthquake Predictability (CSEP), and takes advantage of new large datasets to evaluate <span>whether or not</span> non-linear site effects predicted by site-amplification models are supported by empirical data. The site amplification models are tested using residuals between the observations and predictions from a GMM based only on magnitude and distance. When the GMM is derived without any site term, the site-specific variability extracted from the residuals is expected to capture the site response of a site. The non-linear site amplification models are tested against a linear amplification model on individual well-record<span>ing</span> stations. Finally, the result is compared to building codes where non-linearity is included. The test shows that for most of the sites selected as having sufficient records, the non-linear site-amplification models do not score better than the linear amplification model. This suggests that including non-linear site amplification in GMMs and building codes may not yet be justified, at least not in the range of ground motions considered in the test (peak ground acceleration < 0.2 g).</p>

Sign in / Sign up

Export Citation Format

Share Document